A Conversation about Financial Services of the Future presented by Block | DC Fintech Week 2026
DC Fintech Week·Fri, Nov 6, 2026, 12:00 AM
Join Block for a fireside chat on the future of financial services, moderated by Aaron Klein of the Brookings Institution. This conversation will explore how banking, credit, and lending are evolving to better meet the needs of consumers and small businesses, from new approaches to credit scores and lending decision-making to the role of innovation in expanding access to financial tools. • Moderator: Aaron Klein, Miriam K. Carliner Chair – Economic Studies and Senior Fellow – Center on Regulation and Markets, Brookings • Juan Hernandez, Head of Credit and Underwriting, Block
Transcript
Good morning. This is a headset moment for everybody joining the new technology here and and this is about new technology. This is a conference for the future looking at financial services for the future and I can't think of anybody who's further ahead in time and space than Juan Hernandez who is coming up with a new way to write and allocate credit for society and I'm really excited to learn more about it and have a little conversation about it Juan. So why why don't you start by telling us what it is that Block is doing in creating credit that isn't just simply waking up plugging in a FICO and doing what everybody else does.
Yeah, thanks. Thanks Aaron. Thanks for having me and great to get to chat to you today. Um I'll actually start with like my history at Block. I've been at the company for now 11 years.
Uh and before that, even though you just dissed them, I worked at FICO. Um and I was working at on on the FICO team, I was working on a team that was like kind of half R&D and half like integration of new data sets. And so on the R&D side, we were exploring like can you use machine learning models? Can it outperform like the standard scorecard approach? And then on the new data set side, it was like what other attributes could we be considering to make our models more predictive?
Um and there was a lot of interesting stuff. Um we didn't get that much stuff out into production at the time because we were kind of competing against ourselves and it's hard to compete with the FICO score even when you're at FICO. Um but I came to Block actually uh to work on small small business loans originally. And that was one of the areas that I was focusing on some research on at FICO as well. And the way that we did the loans at Block, this was back in 2015, was we actually underwrote underwrote the loans purely based on a seller's cash flows.
So, this was in the early days if everyone remembers like the Square card readers that you see at a coffee shop. Purely based on that transactional activity, we started giving loans out to customers and we would get the repayments from from that transactional activity as well. And so, at that at that time we realized that there's an opportunity to do underwriting purely on first-party data. What was so interesting about it wasn't just like random first-party data, it was like financial data. So, it was very very closely tied to the repayments that you were trying to get later.
Um fast forward to 2019 on the consumer side, um I was building I was working with a team to build a new lending product for consumers at Cash App. And what we did was a similar approach was based on and let me let me give you like a quick sense of what Cash App is if you're not familiar. Cash App is we're not a bank, but it has a lot of like financial services for consumers. People usually are familiar with Cash App in the context of like peer-to-peer payments where it's very similar to Venmo. Um in addition to peer-to-peer payments, it also has checking accounts, savings accounts, um uh equity investing, Bitcoin investing, a debit card and things like that.
So, a lot of people come to Cash App and they use it specifically to get financial services that they didn't, you know, have access to otherwise. And so, on Cash App we did also consumer-direct to consumer loans. Uh we called the program borrow and the way that we did the underwriting for that was based on all of this activity, we realized, "Hey, you know, you can see people's spending, their savings, you can see peer-to-peer activity, you can see their equity and other assets that they have." And we realized that that that sort of data was like a really, really useful thing to use for underwriting. At that time, and here's like the interesting point, when we first launched the program, we actually used um we used the Cash App data, but we also used a credit report.
And um and we were working with the credit bureaus and we were doing underwriting model. We built an a custom underwriting model that used the credit report as well as the Cash App data. And then after about a year, it was like, "All right, well, we have enough data. Can we try to build a version of this model that uses only the financial data? Doesn't use any third-party, purely data that we have at Cash App?"
And we trained a version of that model. And so usually when you train these models, the first thing that you do is you validate them. And so when we validated it, it came out to exactly the same performance. And so what that means is that the credit report wasn't adding any predictive lift. Um because and I have a hypothesis that we can get into later.
But the the next step after that was "Can we put this model into production and run run an experiment, you know, and like, all right, here's the experiment. Treatment group will be we have a new version of the model doesn't use the credit report. And then here's the control group has the standard model that we have that has a credit report plus our cash flow data." And then it turns out that with the new model that didn't have the credit report at all, we we accomplished what what I consider is is similar to like the impossible trinity and like international like monetary regulation. But basically what we found was consumers were converting at a higher rate.
That was expected because you didn't have a credit check. So consumers would be like, "Oh, this is great. Like there's like no credit check involved." And then the two parts that usually are always at odds with one another was we got higher approval rates and lower loss rates. And the lower loss rates were because the bureaus also if you do a pre-approval, they also apply like knockout rules, they do identity matching and like there's a there's issue with data fidelity that when you do that that route out to a third party.
And so we realized well, actually there's a path forward. Um so that's that's the data that we've been ingesting. Um it's been it's been very interesting to see it evolve. We've been using the data more and more since then. Um we acquired Afterpay in 2022.
Afterpay is a buy now, pay later company headquartered originally headquartered in Australia. It's now part of Block. And we did the same thing. We used the Cash App data for customers that could be um matched with with a certain level of like certainty. We we did a test again, can you underwrite the customer based on the cash flow data rather than the traditional like credit report?
And we found very similar results. And then we just kept expanding that until we realized that there's a There's actually like a building hypothesis here, which is this is probably a better set of ground truth data than credit report data for underwriting customers. And so we just recently um announced the launch of a new product that we're putting out. The idea here is, you know, we've been able to build the out credit access for our customers. Can we just make it portable for them?
And so we've launched a new version of a credit score called the Cash App Score that consumers can take other We're working with external lenders, we're working with other partners and things like that. And the vision here is we have a score that massively massively outperforms like traditional credit reporting and traditional scores for these customers. And we think that there's a path here for like better access to credit for consumers. And we don't offer every type of credit at Block, for example, we don't offer car loans, but this is like a great use case where it's like people who are traditionally not able to get credit, they have like no file, and the only way that they can get credit is by building credit by having credit, which they don't have access to cuz there's always There's always this like chicken-and-egg kind of issue with building credit. Now we just have a score that's based on what they're already doing day-to-day financially, and then using that as a bridge.
And so we've done a lot of validation. We can get into all of the stuff, but that's that's kind of where we're at.
So So there's like three different ways I want to go from this, but let me get back to the comment you made about breaking the trilemma of higher approval and better performance, right? Because there's often this kind of narrative out there that, you know, if you And this was reinforced with the subprime mortgage debacle, which I don't want to take us on a tangent, but like you start giving loans to too far down the system, and they're fundamentally unsustainable. There's this other group saying, "Wait a second, the current system sucks really bad, and if you thought that was a dis to FICO, wait till what we have in store." Coming. And And instead it's the argument is like the current system has a bunch of factual errors and if we could get past this factual error structure one in five Americans is a material error on their credit report.
Mine was an Aaron Klein in New Jersey who didn't pay his AT&T cell phone bill. Couldn't get it off my report even when I said I've been an AT&T customer for 20 years. Here's my bill. No, like nope, nope, not getting off your report. Explain to me or or try to like walk through how much how many more people are you reaching?
How much better was the loan performance? And what were the characteristics of these people that you were reaching that previously weren't served? And what were the characteristics of what products they were repaying cuz you're not right, you're not offering mortgages, right? You're probably not offering $2 million, you know, unsecured lines of credit. So like like walk unpack that a little bit of the success story and then then we'll go to the to the, you know, new roll out world.
Yeah. I think I think you'll kind of hit the nail on the head there, which is the kind of like the false like presumption that people have that by lending to lower credit customers, you're actually going higher and higher in risk. What's really interesting is as we were doing these analyses, we ranked our entire customer base by the traditional scores. And usually what you what you would do like what an underwriter or like an underwriting analyst team will look at is like rank by the score and then measure the default rate by each like 10% score bucket. And so typically what you see is that the the people with the highest score, they'll default at a very low rate and then the people with the lowest score will default at a high rate.
That's how you that's how typically people manage the portfolio. When we ranked everybody by their traditional scores, we found the same default rate across the entire spectrum, which basically just meant that for this segment of customers, the traditional score was pure noise. Like it was absolutely meaningless. And so then when we ranked them with our score, we we were able to actually separate.
What?
What bands were you ranking the we I mean I assume this isn't a 550 FICO to an 850.
We actually So, there's people who use This was on the early data set that we had from Borrow, and there's people that use Borrow across the entire credit spectrum. And so, even the eight the even people in the 800, you know, like people that have 800, some of them will default, etc., right? And so, we actually found that um the FICO score couldn't predict if you were an 800 and you were going to default at all, whereas our score could, just cuz we were looking at higher like higher like think of it as like higher resolution and higher velocity information. And so, like the contrast is and I always I always go back to this example.
It's like if you're if you're measuring for signs of financial stress using a credit report. And it's kind of it's kind of interesting even when you study if you study like recessionary dynamics, like even in a academic level, there's kind of this pre-built assumption that you're going to see credit deter credit scores will deteriorate, and then once you see that, everything starts falling apart, right? And the reason for that is actually like the credit score, it's based on, you know, usage of credit products. And so, you might see a delinquency come in, or you might see an increasing in customers revolving their credit cards rather than paying off the full statement each month. But what's so funny is all of those things, which are like the bedrock of our understanding of finance from like an academic perspective of like what happens in credit credit-based recessions, it's actually you're you're already starting from like the wrong data.
When you see a default on a credit card, you're already seeing like you would have seen that much much sooner looking at cash flows. So, think about like a Think about it somebody who like lost their job or something like that, or had like a massive unplanned expense. You're going to see an imbalance in their cash flows. You're going to see like either a shock to inflows or a shock to outflows. And then maybe a month later, maybe 2 months later, you'll see them slip somewhere in like how they use credit products.
And then a month after that, after it gets reported back to the bureaus from the lender, then you're you're to see it change to the FICO scores. And so, that's the part that's like when you think about it this way, it's kind of shocking that like we actually have a very fragile system where everything is based on a score that's inherently very, very lagged.
So, uh uh I was an orchestra dork. And I played the oboe. And anybody here an orchestra person, right? So, you would tune to the oboe. When the orchestra sits down, the orchestra tunes to the oboe.
Now, why does it tune to the oboe? It's not because the oboe produces the best A. The violin produces the best A. The purpose of the orchestra is not for each individual person to try to make the clearest note. The purpose of the orchestra is to sing in tune.
And unless you have perfect pitch, you are not going to be able to know if the entire orchestra is playing a 438 for an A instead of a 440. That's the frequency if you get all scientific about it. The purpose of the financial sym- symphony is not to maximize default risk. It's not to be the best, most efficient use of capital. It's to make the most profit where underwriting is expensive, and, you know, i- i- i- the structure of this.
The purpose of the credit scores is not to get the most accurate credit FICO number for everybody. It's expensive to figure out all the Aaron Kleins. Let me tell you, I've gone down that rabbit hole.
Or all the Juan Hernandezes.
I mean, God for- You're You got it worse than me. And so, the purpose of it all is to be in tune. Now, you guys are staying out of tune a little bit. You're producing this block score, and my question is like, I You had me at go in terms of like the idea. And by the way, it's not just your work.
Cash flow underwriting, FinReg Labs has a paper, Manju Puri from Duke has a paper. There's a other people who have come across this reality. A- um But you're actually implementing it. How does it work now? So, you know, I have this crappy FICO because it's all messed up and jacked and I have a better block score.
Like what do I do you know like I it's awesome that I can like can I go take it somewhere and show it to the car dealer like how does it work?
Yeah, that's that's how we want it to work essentially. It's going to take time before we build out all the partnerships. Um but basically let's I compare a customer in like two ecosystems. One of like let's say they're like a new new immigrant to the country, right? They come they're established.
Um they've got a job. They have cash flows. They don't have credit. And so typically they're a no hit. If they want to get credit, there's a there's some services now that like try to help you but like let's just play out the example like kind of a little bit more old school.
Um if they want to get credit then they have to like get a secured card or take out some sort of debt instrument with like a cosigner and then maybe over years they evolve from like no credit to like a 500 and then like a 600, right? But like all the while let's say they have the same job. They bring in their savings from you know wherever they came from and then like they're gradually so on path A which is the traditional path they're like very very gradually getting to like 500 and then the next year maybe 600 and then eventually to like a 700. Let's say their true risk is the equivalent of like a 790. I'm just picking a random number.
Um that's what you would see in day one if you're looking at their cash flow data because the cash flow data is like they're using their cash flow the exact same way month over month. They're paying the same bills. They're building their savings at the same rate. They they manage like their discretionary spending and it's like necessities and the same way. And so the difference that you would see is you would have you would get to that 790 figure sooner and and the the point that it we're making is that the 790 number was true the whole time.
It's just that the way that you build out that that score in the traditional way is like is very gamified like in it's gamified in like this very negative way which it's I don't maybe maybe a better word for it is like it's performative. Um you know I'll I'll give you like examples from my own life, right? It's like because I worked at FICO, I know the the dumb tricks that you have to do, right? It's like, okay, well, I got this credit card I never use, but it was my first credit card ever. I'll just keep it open and like use it for like a coffee every like few months or whatever.
It's It doesn't make any sense, but that is actually a signal that if I close that account, my FICO score would go down. Or if I travel for work, occasionally I have to expense something on my other card, which will drive up my utilization and then my FICO score goes down instantly. But it's like a very recurring regular thing. Nothing's changing in my underlying risk.
Yeah, I mean it's You told the Yelena McWilliams story, right? We all Everybody here heard hear Yelena tell that story over the course of her tenure, right? It's the exact same uh uh uh story that she was a spectacular credit risk and went through all this product. And the gamification And the gamification works the other way. Those of you guys may know there was a national bank that used to change the data they reported back to the bureau on their customer profile, not give the maximum amount of spend cuz they could they could make their customers FICO look worse so they wouldn't get poached with credit cards.
So, as much as we you may think you're playing the game or I'm playing the game, like there're people getting paid real money to play the game against you. How are you going to be able to break this game and cycle that we're all trapped in and provide better opportunities for for for the Yuzus of the world, for the Yelenas of the world, for the Aaron
Kleins who can't get the bad Aaron Kleins up off of their credit report of the world?
Yeah, that's a great question because yeah, you always wonder like, well, can't people gamify the Cash App score? The difference is the Cash App score is based on like what I call more fundamental sources of information. So, like your savings rate, your discretionary spend, your cash flows, your income, your other, you know, like how you're actually spending month to month. And so, in a sense, it's not falsifiable. You can't like You could put more money into the system, but that means you have more money.
Like that that That's exactly the kind of signal that we're looking for. So, that's very different from okay, a person who for whom it makes no difference, you know, regarding their level of risk, you know, they just they just learned credit building. They like maybe they had to repair their credit at one point and then they learned the tricks or whatever. It's like that's a very very different thing. Like you're not you're not suddenly you don't suddenly have a higher income, you don't suddenly have higher net worth or anything like that.
And so, you can't really falsify these things.
It's looking at the mirror, right? Like like you know, with with your clothes off. Like you you you you can't hide your financial history when you're looking at your bottom line. It's one of the reasons I love it, particularly as my research has shown, that as people have structurally gotten more and more different sources of income. And I'm really glad you used the shock to income and not the shock to expense because for most of these folks it's actually variable income that's driving problems.
In Washington you hear the story about variable expense and you always hear the same story. Right? Car broke down. Right? That's actually the number two expense shock.
The number one expense shock is funeral. Why does Why does nobody talk about funeral? Because all of us pay for our own funerals. We don't live in the world of people where you get a call, your aunt died, and you are now on the hook for a thousand bucks to pitch in for the funeral, and you're not going to delay that, and you're not going to say no. Right?
But that's an uncomfortable conversation, doesn't resonate with hill staffers, certainly doesn't regulate resonate with people in the bank agencies who have very steady incomes, right? They they just know about expenses. And this is why the agencies constantly get it wrong because they're utilizing for solving for people like themselves, which is not the population of folks you're you're dealing with. You can't hide. If I have a second gig, if I have it like like your structure is better for the future of work because I can't hide that if I'm getting if I have a side hustle, second gig.
I mean, the how many different streams of income are you seeing on these people?
Yeah, that's that's really interesting and probably like a good segue into a topic here which is so so to answer your question like we see I can't remember the exact percentage of customers but we see a large number of customers that use Cash App have multiple sources of income and so part of the reason that you guys might not have ever used Cash App day-to-day is for the exact same reason that Aaron was just describing. Cash App has really made inroads in parts of the country that have like lower income and things like that. Again, the reason for that is we started with peer-to-peer. That was the original technology that we had done. I remember the first day somebody sent me $5 via email when we first started Cash App and I was like, "What?"
And then you can send it just by writing in the subject. That doesn't seem right. But
[cough]
but so the cool thing is when we built this
[clears throat]
and we added more features to it, this is the part that's very different from Venmo where me and you will use Venmo and it's like all right, I use this thing to send you money. I'm never going to keep a balance. I'm never going to look for financial services. But we actually found like a fit with a with a demographic and a consumer that was like, "Wait, actually my bank doesn't give me free ATM fees when I'm out, you know, out of using an out-of-network ATM. We do that.
My bank, because I'm like a 500 FICO score, they're going to charge me $25 for overdraft. We don't charge anything for overdraft. And so we basically found that there's a whole population of customers they want to use Cash App but then there's a whole other, you know, that's like the tip of the iceberg. The other thing is like, "Well, how do we make the system work better for them?" And so for for these customers, it's just been like a huge learning of like one learning I already described which was like, you can have a customer that has a 580 FICO score.
In fact, 70% of customers who use borrow, which is our lending product, have a 580 or below FICO score. That's like 70% of the portfolio.
And what's their default rate?
It's it's like super low. So historically like 3% and like we've we've been managing it to like our target. So it's always been at our target level. And so what's crazy about this is like when we tell you know, like if if we if I were shop if I were going around shopping for capital on like Wall Street or whatever is like 580 FICO score for 70% of the portfolio, I'd be like laughed out of the room typically. But, this is the part that like you have to look at the data, you have to look at the evidence, and you realize, "Wait, actually the the premise that 580 is like bad is is what's wrong about the equation."
Well, I mean that that's that's part of the whole problem with the the financial ecosystem and the over-reliance in this very outdated, very inaccurate structure uh is based on that. And the fact of the matter and this is what I think is so interesting about the product is that you're reaching this customer base, right? Like you talked about identity earlier on. So, anybody here know second chance checking? Anybody here on the do not bank list?
About uh uh First of all, nobody knows how many people are on the Chex do not bank list cuz Chex won't tell you, but at one point there was one estimate of 5 million comes from Chichi Wu's paper, got a little momentary piece of data. So, this do not bank list is used for AML verification by many banks. It has nothing to do with whether or not your money laundering risk. It's simply did you leave a bank previously with a negative balance? Which often happens when you leave college and you close a bank account, and then you missed a minimum monthly payment, and they charge you a fee, and then the fee went on and went to a mailing address, and you never got it, and you don't even know about this thing, and you're on the list.
A lot of these folks people tend to go have problems. They got a divorce, they got a bankruptcy, something happened. They had a problem. They've now closed that chapter in their life. America's built on second chances, and they're they're doing awesome, but they're on this list.
And so, they can't get a bank account. So, then you get these horrible banks like Woodforest that goes around and just overdraws the crap out of people. They average $94 of overdraft fees per customer. Bank of America averages less than two. Right?
Why? Because they target these customers. 8% of Americans are responsible for 80% of the overdrafts. Those folks are paying five, six, seven thousand dollars a year for basic banking services. And then you come along and you're like, "How much does it cost?"
Uh nothing. Yeah, there's no annual fees for banking services on Cash App.
Right. And then your overdraft?
It's free.
Yeah. And can I can I borrow 500 bucks?
Yep.
[laughter]
So this is I mean like this is this is the idea though. And and so as you can see like the the scoring piece of it is like part of it. The other thing is um it's consumers who've been like really taken advantage of and has created demand for totally new types of products. That's the other like big thing that's been evolving. Um it's not I mean like it's not brand new anymore, but like buy now pay later is like an example of this where there's a lot of consumers who, you know, they're used to if I get a credit card is like 30% APR, they want me not to pay so they have a super low monthly minimum and then like I'm going to end up paying more in interest than I originally borrowed over time.
Like these are the types of customers that are coming to Cash App and are using new forms of credit as well. So there's been like an explosion in the last 10 years of different forms of short duration credit that are usually almost almost always like very closed ended like closed off where BNPL's an example where Afterpay um is Afterpay is a block BNPL, but there's also like Affirm, Klarna, and others. They basically offer like an an installment plan per purchase. And so that's like a thing that people have come to cuz they like it because it's like, "Okay, six weeks, I'll pay it back. It'll be done."
There's also um earned wage access which is basically This is a really interesting one because the way that this one works is if you know people's income, you can give them an advance on that income if they're if it's coming in through payroll. A lot of them work by intercepting payroll and so there's no consumer fee because there's very little loss. A lot of customers really like that because like what they're solving for isn't necessarily what we're solving for. They're solving for for example, I don't know, funeral's a good example. It's like I get paid, you know, next Friday.
I get paid on the 3rd of the month.
Yeah.
I get paid on the 3rd of the month. By the way, the money's taken out of the employer on the on the 28th, 30th, right? But my my bills Oh, you know what? I'm going to I'm going to delay all my auto pays. And by the way, if I'm 2 days late on my auto pays, what happens to my FICO?
Yeah, exactly. Yeah. And with more utility reporting into FICO, then you get more and more screwed with that, too.
Right. The things people think are solutions are actually exacerbating the problem because we haven't addressed the root cause of the problem. We're just papering over it because the system is being run by people who have never experienced life in that room. Uh and they think, you know, so we got 2 minutes left. What's the future of the product?
Cuz you you described taking the score. How are you going to break into an orchestra that's deeply out of tune with a brand new beat?
Okay. So, I think one of the ways that we're very different with our score compared to a bureau is as I mentioned, we have a we have a we have a banking relationship with the customer. Nobody is like logging Nobody like Like Experian has never heard of you outside of like owning your score, right? Like this is a thing that was forced into existence with no relationship with the customer. We're trying to evolve with the customers and serve their needs.
And so, we're interested not just in can you build a score that's like predictive, but can you build a score that people can like take action on and have like real tangible benefits for them. So, we're working actually um with um with with lenders that operate at our target customers. And then through distribution as well, where customers I'll give you an example like some of the things that we're looking into. Uh auto loans already came up earlier. We did a study similar methodology to like our model scoring study that that I mentioned earlier.
But with car loans, we measured like a third I think it was like a 30% I could be slightly off, but it was like a 30% higher approval rate for car loans at a same target loss a traditional like lender would would usually have on their book. And so, an example that I always come up with in my mind is like imagine your your typical Cash App customer, you might live paycheck to paycheck or maybe better than that, but you might have unplanned expenses or you might have variability of income and things like that. You might need a car for a new job and under the traditional system, you have nothing. But you could actually just see, for example, you're pre-approved if your Cash App score is above X, you could see it in the app. We could partner with lenders to make those offers available in the app so you could see what access your building over time or you could do it directly through the lender as well.
So, it's that. Rent screening is another example I could get into, save $75 when you're applying for rent. It's insane that customers have to pay $75 at each at each property, right? That's completely insane. You should do it once and then never you should never pay for it to begin with, but you should be able to reuse it.
So, this is again, it's portable, it's in the customer's hands.
Well, that's fantastic and exactly on time. Join me in thanking one for this fantastic conversation.
[applause]
Thank you, Aaron.
And uh look forward to seeing everybody's Cash App score.
Great talking to you. Thanks.
Automatic transcript — names and jargon may be misspelled.